XNAT: A SOFTWARE FRAMEWORK FOR MANAGING NEUROIMAGING LABORATORY DATA

1
XNAT A SOFTWARE FRAMEWORK FOR MANAGING
NEUROIMAGING LABORATORY DATA D.S. Marcus1,2,
T.R. Olsen1, M. Ramaratnam2, R.L.
Buckner1,2 Howard Hughes Medical Institute1,
Washington University in St. Louis2
INTRODUCTION
XNAT in operation
The XNAT schema defines experimental data and is
used to generate the archive components.
XNAT manages a terabyte-sized neuroimaging
archive at Washington University. Over 1800
imaging sessions from the Universitys
Alzheimers Disease Research Center, Conte
Center, and Cognitive Neuroscience Laboratory
have been entered, vetted, processed, and shared
in the archive. Data include structural MRI
images, DTI, research radiological evaluations,
white matter integrity assessments, automatic
manual volumetric measures, and normalized whole
brain volumes.
Neuroimaging research generates large quantities
of raw, processed, and derived image data.
Studies often include related non-image data,
such as demographics, genetics and clinical
measures, as well. Maintaining the integrity,
organization, security, and accessibility of such
data requires substantial software
infrastructure. Additionally, informatics tools
are needed for exploring and mining the data. We
are developing a toolkit the Extensible
Neuroimaging Archive Toolkit (XNAT) based on
industry standards to support these requirements.
XNAT uses XML Schema to define experimental data.
XNAT automatically builds a database backend for
storing data, while an accompanying web
application allows users to archive, manage,
search, and visualize the data.
Database definition
CREATE TABLE table ( name, VARCHAR(50),
idMethod, VARCHAR(50), DEFAULT 'null,
type, VARCHAR(50), ID INT UNSIGNED, NOT NULL
AUTO_INCREMENT, PRIMARY KEY (ID) ) CREATE
TABLE table_column ( name VARCHAR(50)
javaName VARCHAR(50) primaryKey VARCHAR(50)
XML Schema
Java classes
lt?xml version"1.0" encoding"UTF-8"?gt ltxsschema
xmlnsxnat"xnat.xsd" xmlnsxs"http//www.w3.org/
2001/XMLSchema" elementFormDefault"qualified"
attributeFormDefault"unqualified"gt ltxselement
name"Database"gt ltxsannotationgt ltxsappinfogt ltxna
tdbinfo database.name"CNDA" database"MYSQL"/gt lt
/xsappinfogt lt/xsannotationgt lt/xselementgt ltxsel
ement name"wm_mod_scheltens"gt ltxscomplexTypegt ltx
ssequencegt ltxselement ref"experiment"/gt ltxsele
ment name"session_id" type"xsstring"/gt ltxselem
ent name"assessment"gt
public class Experiment extends
org.cnl.cnda.om.BaseExperiment implements
Persistent public static final long
MILLIS_IN_DAY 86400000 public static long
getDays(Calendar c1, Calendar c2)
long time1 c1.getTime().getTime()
Data file format
lt?xml version"1.0" encoding"UTF-8"?gt ltsession
id"050202_vc1000"gt ltsubject map_num"0001"
lab_id"0001"gt ltagegt78lt/agegt
ltgendergtmalelt/gendergt lthandednessgtrightlt/h
andednessgt ltcohortgtADRClt/cohortgt
lt/subjectgt ltsession_infogt
ltacq_dategt5/1/2002lt/acq_dategt ltpigtRandy
Bucknerlt/pigt ltoperatorgtDan
Web app
lt!DOCTYPE html PUBLIC "-//W3C//DTD HTML
3.2//EN"gt page.setTitle("CNDA -- Integrating
the Neurouniverse") page.setLinkColor(ui.alink)
page.setVlinkColor(ui.vlink) lthtmlgt ltbodygt ltta
ble border"0"gt lttrgt lttd bgcolor"336600"gt

Quick Search locates subjects and displays a
dynamic summary report.
Add experiment link displays data entry forms.
METHODS AND MATERIALS
XNAT utilizes a number of supporting technologies
and software packages. In all cases, open source
and/or freely available libraries have been
selected to ensure that XNAT is a zero cost
solution. The specific packages are indicated
below. A number of aspects of the XNAT
architecture are in beta or earlier production.
XML Schema XNAT schemas follow the XML Schema
1.0 specification. Of note, XNAT utilizes
xsappinfo tags to incorporate custom application
directives. XNAT includes a base schema that
includes elements for representing common
neuroimaging data. The base schema can be
extended to capture additional data types. Data
storage XNAT operates on data stored in a
Postgres relational database. XNAT automatically
generate the database tables from supplied XML
Schemas. XML instance documents can be loaded
into the database using XNATs command line
tools. Alternatively, data can be entered via
custom forms within XNATs web applications. All
image data are stored in binary files. The base
XNAT schema includes elements detailing raw and
processed image files and their locations on the
server. Web application The web application uses
the Jakarta Turbine platform , which provides a
user authentication model, data object
generation, and Model-View-Control development
architecture. The Velocity templating language
is used for creating web pages. PDF generation
Dynamic PDF documents are generated by
transforming XML data documents into XSL-FO
documents which are then passed through Apaches
FOP engine. Image viewing The XNATs online
image viewer is a custom Java applet based on
ImageJ. Image data is transported from 8-bit
image files to the applet using HTTP
tunneling. Command line tools XNATs command
line tools are written in Perl (using the
XMLDOM and DBI modules for XML and database
operations) and Java.
The XNAT architecture includes clients,
middleware, and data storage.
Client applications
Experiment links generate reports and action
lists tailored to user and data type.
XDAT engine
XDAT API
XNAT Schemas
Form handler
Search engine
List generator
Database generator
Report generator
Dynamically-generated PDF documents serve as the
hardcopy medium.
XFT
Data store
Relational database
Data files
The web-based image viewer allows users to view
raw and processed images.
XNAT is a framework for laboratory data
management and sharing
XNAT improves data security and data integrity.
XNAT
SUMMARY
XNAT provides laboratories with a framework for
managing their data. Experimental protocols can
be captured using XNATs default schema or a
custom schema. The schema is then transformed
into a number of customized tools for quality
control, processing, and exploring. XNATs
extensible design allows laboratories to quickly
capture new experimental protocols by simply
appending their schema documents, ensuring the
long term maintainability of their data archive.

• Provides a single, integrated data store.
• Enables quality control measures.
• Provides links between related data types.
• Includes online management and exploration tools.
• Provides user access control.

• Data is easily lost and mishandled.
• Data sharing is difficult.
• No way to link related data types
• Difficult to control user access.

• Password protected access.
• Create, read, update, delete privileges granted
  by lab experiment type.
• Auto-logoff from website.

• Data entry validation.
• Automated processing pipeline.
• Database backup to XML.
• Automatic quality control email reports.

This work was supported by NIH-BIRN, NIA
(AG00030, AB03991), NIMH (MH57506), the
Alzheimers Association, the James S. McDonnell
Foundation, and the Howard Hughes Medical
Institute. Contact dmarcus_at_wustl.edu